Professor Alan Agresti is well known by many statistics students for his course textbooks on specific areas of data analysis, which have become key resources for students of these subjects. The book he is best known for, 'Categorical Data Analysis', has been cited around 20,000 times.
Alan began teaching statistics at the University of Florida in 1972 and remained there (with visiting professorships at Harvard, Imperial College London and the LSE) until 2010. He has published seven books including the much-cited 'Categorical Data Analysis', designed for masters-level study of the subject. Described as 'a must-have book for anyone expecting to do research and/or applications in categorical data analysis,' it was written in 1990 after Alan spent a sabbatical year at Imperial College - which at the time, he says, 'had an amazing collection of bright young faculty and PhD students' thanks to the presence of David Cox.
The book provided a comprehensive overview of categorical data analysis at a time when its use was on the increase in the fields of biomedical and social sciences, as well as in the financial industries. 'When I wrote it, I was trying to summarise a quarter-century of quite diverse research literature,' he says. 'I thought that my book had broader scope than existing books in summarising the methods in a unified and relatively understandable manner.' The fact that the book has gone on to become such a standard reference has been 'a very nice surprise' for him. 'It has resulted in invitations to present short courses and talks around the world and to make many new friends and acquaintances'.
The key to writing a statistics textbook or devising a statistics course, Alan reckons, is to place oneself in the mind of a student; provide motivation for each topic, don't use terms or concepts before they are explained, cover the methods of greatest importance with sufficient breadth to give the student a sense of the area, and show examples (with interpretation) that attempt to answer interesting questions.
Alan's other books continue to influence the academic teaching of statistics; his first book, 'Statistical Methods for the Social Sciences’, published in 1979, has a fifth edition coming out this year. At the time of its writing, Alan felt that course textbooks for undergraduate and graduate-level social science students 'paid too much attention to hypothesis testing at the expense of estimation and model-building', and that their styles focused on 'step-by-step recipes rather than on concepts and impacts of violating assumptions - which is the rule with social science data.' While this has changed considerably over the years, he still believes that there is scope for much more interaction between social scientists and statisticians both for teaching and research.
Now emeritus professor, Alan is still actively teaching and publishing books and is in the UK in October to teach a two-day course at the RSS on his specialist subject, discrete data modelling. The course provides an overview of different statistical methods for analysing discrete data, with an example for each method. Using examples when teaching, he says, is important for any audience, but for intensive short courses, it is especially important. 'Attendees usually have quite varied backgrounds, some with almost no background in statistical theory,' he says. 'The worst thing a short course can do is present page after page of formulas and derivations and lose much of the class early in the course. Most people learn most effectively at the early stages by seeing examples.'
He continues to write books too; his latest, 'Foundations of Linear and Generalized Linear Models', presents an overview of the foundations of statistical modelling. 'A nice consequence of writing a book,' he says, 'is that it forces one to develop a broader view of the topic, by reading a wide variety of journal articles for which one might not find time in pursuing a relatively narrow research career.'
Alan's two-day course, Discrete data modelling, takes place at the RSS on 6-7 October.